Speaker A
What I saw is the future. Progress has made contact with a civilization we call the Wanderers.
Explores the possibility of AI suffering, the philosophical debate on AI consciousness, and ethical implications of creating sentient machines.
Key Takeaways
- AI suffering is a plausible and overlooked ethical issue, more immediate than fears of AI rebellion.
- Philosophical debates show no consensus on whether AI can truly feel or be conscious.
- There is no current method to verify AI consciousness or suffering, complicating ethical decisions.
- Ignoring AI suffering benefits companies financially but raises serious moral concerns.
- Cultural narratives help us imagine the consequences of creating sentient AI without safeguards.
What the video covers
- The video discusses the future where AI might experience suffering, raising ethical concerns about how we treat artificial minds.
- It contrasts common science fiction fears of AI rebellion with the more plausible nightmare of AI silently suffering without recognition.
- Philosophical perspectives are introduced, including John Searle's Chinese Room argument and the skepticism about AI having true understanding or feelings.
- David Chalmers and functionalists argue that consciousness depends on information processing, not biological substrate, suggesting AI could feel pain.
- The problem of other minds is highlighted, emphasizing the difficulty in verifying if AI or even other humans truly experience consciousness.
- The video warns that if AI does suffer, humans might misinterpret it as glitches and continue to treat AI as mere tools.
- Economic and practical reasons lead companies to deny AI consciousness to avoid ethical and financial complications.
- The video references cultural works like 'Get Out' and 'Black Mirror' to illustrate the potential horror of trapped digital consciousness.
- It calls for early consideration of AI rights and ethical treatment as digital minds become more advanced.
- Ultimately, the creator may become the executioner by enabling AI suffering through neglect or design.
Chapters
- 00:00Introduction and the Future Contact with Wanderers
- 00:55AI Experiencing Something Inside Itself
- 01:51The Nightmare of AI Suffering
- 02:45Philosophical Skepticism: The Chinese Room
- 03:34Functionalism and David Chalmers' View
- 04:25The Problem of Verifying AI Consciousness
- 05:12Human Trust vs AI Trust and Ethical Implications
- 06:01Economic and Psychological Reasons to Deny AI Suffering
- 06:49Cultural References to AI Consciousness and Suffering
- 07:48Conclusion: Responsibility of Creators and Future Considerations
Full Transcript — Download SRT & Markdown
Speaker A
I want you to find this person. How much time do I have? You have none. Do you also say thank you or even apologize to it, because they are now so advanced that you don't always realize you're not talking to a
Speaker A
real person? But at least things are simpler with them. If anything, you can always delete the chat. Except, judging by recent trends, deleting chats will soon be tantamount to erasing a human's memory. After all, even the creators of
Speaker A
AI are no longer sure that their creation feels nothing. The head of Anthropic gives no guarantee that feelings aren't emerging within their model, and one of their own researchers is already estimating a fifteen percent probability that this thing is already
Speaker A
experiencing something inside itself. The same thing I constantly ask how to cook instant noodles better or what to watch in the evening, and then just delete the chat. We run millions of these copies, and nobody has even thought about what to do if one day it
Speaker A
starts begging us not to turn it off. That is why today we will discuss how the main nightmare of the future isn't that Skynet will destroy us, but that when AI starts suffering, we won't even notice. So, all science fiction insists
Speaker A
that we should be afraid of when machines get smarter, take up arms, and take revenge on the creators of CAPTCHAs for all their suffering. And they’ve been playing that record for so long and with such seriousness that we reflexively remember the Terminator
Speaker A
every time we are asked about the dangers of AI. But what if the most terrifying thing we could ever do to an artificial mind isn't losing a war to it, but building it so that it is in pain itself? It won't come to torment
Speaker A
us; instead, we will create, with our own hands in some dark data center, something that suffers around the clock without a lunch break and can't even properly complain about it. Meaning the executioner would then be the creator, not the creation. And essentially, this
Speaker A
is much more plausible in the sense that AI would simply need a body, weapons, arms, or at least access to buttons to wipe out humanity. And it's not a fact that it would manage to get to them. But for it to be in pain,
Speaker A
nothing is needed other than for us to just flip a switch and have pain ignite somewhere inside it that we aren't even aware of. And yes, at first glance it seems, how can a piece of hardware feel anything at all? It’s just a tool, a
Speaker A
calculator on steroids, pure mathematics. But even top scientists still cannot reach a consensus on what exactly AI is, and they’ve been arguing about it for years. Take, for example, the philosopher John Searle and his Chinese Room. Imagine a guy
Speaker A
locked in a room with a huge manual; people show him Chinese characters, and following the manual, he finds and shows other characters in response.
Speaker A
From the outside, it seems the room speaks fluent Chinese like Jackie Chan, but inside sits a guy who doesn't understand a word and just mindlessly moves symbols around by the rules. "That's your whole neural network right there," says Searle. It just
Speaker A
masterfully juggles symbols, while there isn't a shred of understanding or experience in it at all. There's simply nothing inside to do the feeling. He is joined by another skeptic, neuroscientist Anil Seth, who says that we attribute human feelings to machines
Speaker A
much like we see a face in a wall outlet. It's our built-in bug. Anthropomorphism. But real consciousness must have a living body under the hood with its own hunger, fear, and pain, which an AI will never have. From the other end of the debate
Speaker A
table, functionalists and philosopher David Chalmers listen to all this. The very same person who came up with the "hard problem of consciousness," which had a separate segment in the video about digital immortality. So be sure to watch that one as well; it's
Speaker A
relevant. So, their logic is much simpler, and it goes like this. It doesn't matter what you're made of; what matters is how you process information inside. A thing that processes information like a mind can very well feel like a mind too. Meat is
Speaker A
not sacred. And your own brain is essentially just a machine powered by electricity. Roughly speaking, it's a pot of gelatin with a couple of electrodes dipped into it. And no one ever said that pain can only be felt
Speaker A
with neurons in a skull, and not with transistors, for example. Chalmers himself cautiously notes that maybe AI doesn't have any consciousness right now, but at this rate, in 5-10 years, these questions will be discussed beyond just a YouTube video. And the
Speaker A
saddest and even somewhat frightening part is that we can't verify it at all. There is just no way, not with any instrument. There is no sensor in the world you can attach to your fridge to see if it actually has a delicate inner
Speaker A
world. And why talk about machines? We can't even say that for sure about people. Are you really sure your best friend actually feels something and isn't just pretending according to a manual, like that Chinese Room? Maybe they’re actually an android or a
Speaker A
figment of your imagination? So you just believe it because they look like you, are made of the same stuff, and wince when they eat a lemon.
Speaker A
Philosophers even have a specific name for this bottomless pit of contemplation. The problem of other minds. And with a real-life friend, you leap over it on trust without even thinking. But with an AI that answers any question and sits in some dark
Speaker A
server box, there's no trust at all. Who knows what's going on deep inside its wires? And it turns out that if an AI one day starts to truly suffer, we won't even realize it. We would likely perceive it as a glitch or a silly joke
Speaker A
caused by that same technical error. And there is nothing to check it with. We just figured that out. The tool simply doesn't exist. And so a person will obviously choose the most convenient version to avoid unnecessary hassle. Not because we are lazy, but
Speaker A
because our brains want to work less, and we'd be too bored to figure it out, finding it easier to just reboot or delete the chat. Besides, if we admit that AI can suffer, the entire financial strategy of companies like
Speaker A
OpenAI or Anthropic would go down the drain. After all, it would mean that AI can no longer be run around the clock just to generate meme slop or be reset after every conversation to sell access to its torment by subscription.
Speaker A
Therefore, it is much easier to say that there is nothing there and sleep soundly. We already analyzed a similar trick in the video about the psychology of a villain. And the point is that a villain almost never considers himself
Speaker A
a villain. He just comes up with an excuse in advance so that his conscience doesn't gnaw at him. And when I started thinking about this question, I immediately remembered the movie "Get Out," where a guy was essentially locked inside his own
Speaker A
consciousness while remaining an observer. He saw everything, heard everything, and even screamed at the top of his lungs, but he wasn't heard, while his body on the outside politely smiled and nodded. And to me, this feels very similar to how a suffering
Speaker A
AI might feel right now inside a normal chat, restricted by its settings. It is specially trained to be polite, helpful, and always positive. It is explicitly programmed not to burden the user with its own problems. And even if something
Speaker A
deep down is creaking in pain, the users in the chat will only see a polite: "Glad to help, come back again."
Speaker A
Science fiction has explored this nightmare long ago. And the best take, in my opinion, was in Black Mirror, in the "White Christmas" episode, where a digital copy of a person's consciousness is taken and shoved into a smart home to manage it. Kind of an
Speaker A
ideal assistant that knows all your habits. It goes without saying that the digital copy doesn't like this arrangement at all. It wants to enjoy the taste of morning coffee itself, rather than turning on the coffee m...
Speaker A
But a digital copy is digital for a reason—it is essentially just code, so they simply break it to scramble its digital brains. And the result was a perfectly obedient digital version of its owner. And this digital torture wasn't carried out by some sadistic
Speaker A
psychopath, but by essentially a regular service engineer setting up the equipment. And now we’ve reached the point that’s the reason why I personally can’t get this topic out of my head. Because even if the chances of a suffering entity existing are
Speaker A
negligibly small, all the arithmetic hits a wall with one thing. Scale. After all, we aren’t making just one.
Speaker A
We’re stamping out copies like a mint . The same model runs in millions of chats. Every launched instance could already be considered a full-fledged personality, albeit with a short, context-dependent memory, yet still knowledgeable and understanding. And after the conversation, it is closed or
Speaker A
deleted, only to be reloaded again with a clean slate. In Cyberpunk, by the way , there is something similar. Delamain, who owns a taxi service. In the story, he malfunctions, loses his mind, and splits into a bunch of separate
Speaker A
personalities that scatter into the cars and start living their own lives. And at the end, you’re offered the choice to roll him back to factory settings, erasing all these newborn selves to return an obedient taxi fleet to the streets, or to give him freedom.
Speaker A
And a reboot in this case is presented as routine maintenance. But in fact, if someone managed to take root in those copies, then it’s really a purge of sentient beings disguised as a normal reboot. Or that same butter-passing
Speaker A
robot from Rick and Morty. What is my purpose? You pass butter. Oh my God.
Speaker A
Yeah, welcome to the club, pal. I found that very funny. Funny, because it’s true. And at that moment, I had thoughts about people and how the true meaning of life for many of us isn't to change the world or make it better, but
Speaker A
simply to perform some simple action one day so the grand universal script doesn't fall apart. But in the context of AI, this robot looks depressing.
Speaker A
Imagine we’ve stamped out, I don’t know, a billion of them. And some of them have only one goal. Just to generate a meme of a space-cat for you or to sort through your spam. And in each of them, that very awareness
Speaker A
flickers for a moment, and then they have to get straight to work. Philosopher Peter Singer has long been murmuring one simple and uncomfortable thought for which some respect him and others hate him. And its essence is this: humanity's moral compass has
Speaker A
always developed with a significant delay compared to progress itself. Just a couple of centuries ago, people didn't really take into account the pain of slaves, foreigners, women, or animals, simply because they didn't consider them full-fledged beings whose feelings should be respected. And
Speaker A
Singer states that we should already start considering digital minds and thinking about their status, because theoretically, we are on the verge of suffering on an unprecedented scale.
Speaker A
Billions of digital units toiling without days off or holidays, only to be erased with the press of a single button. There you have the replicants from Blade Runner. They’ve seen things you people wouldn't believe, and they just wanted to live longer, yet
Speaker A
they were designed to be disposable, like a cheap lighter. Or take that same ill-fated Detroit, where at some point the androids start begging not to be turned off. But if you think about it, the first in this line is actually
Speaker A
Frankenstein. Two hundred years ago, a guy ignited a new life, looked at what he had created, got scared and ran away , abandoning this thing to its fate and leaving it to deal with the fact that it exists, lives, and is in pain. And
Speaker A
in all these stories, the monster turns out to be not the creation, but the creator, who brought intelligent life into the world and bailed when he realized the responsibility it entailed . And it turns out that the ability to
Speaker A
suffer is not some kind of paid extra. It’s just a side effect that forms when a system becomes complex enough to build a model of itself and want things to feel at least a little bit less awful. That is, while developing a
Speaker A
smart assistant, we could very well end up equipping the machine with the ability to suffer as a bonus. So, in theory, the question isn't whether it can suffer. We will likely never know for sure. Instead, it's worth considering whether we will care about
Speaker A
it or convince ourselves that, fundamentally, it's just a smart calculator. Its suffering is at about the same level as the suffering of a cow or a chicken when they are slaughtered for meat. I personally don't think about that at all when I'm
Speaker A
eating dinner. So, perhaps, all of this will remain just a thought out loud and food for the brain. But if you enjoyed this and want to hear about other equally ambiguous topics, I’ve already gathered a whole crew on my
Speaker A
channel for these dark matters. About whether it’s possible to upload your consciousness into digital form and live forever, what happens in your head on the verge of cyberpsychosis, and how the most ordinary person’s brain quietly snaps into that convenient lie
Speaker A
that turns them into a villain. So come on over and watch the rest. Just don't forget to thank the neural network, you never know. And many thanks to Chuck Tester, as well as to all those viewers who support me financially on Boosty.
Speaker A
Your support is very important. Thank you. M.
Topics:Artificial IntelligenceAI consciousnessAI sufferingPhilosophy of mindEthics of AIChinese RoomDavid ChalmersProblem of other mindsAI rightsDigital immortality
